The Agent Skills Directory

[DATA_EXFILTRATION]: The skill fetches external data from PubMed and PubMed Central (PMC), which are well-known and trusted scientific databases. This behavior is essential for the literature review functionality.
[PROMPT_INJECTION]: There is a potential surface for indirect prompt injection. The skill ingests untrusted external content (full-text PDFs and abstracts) and passes this data to sub-agents for analysis. While no explicit malicious instructions are present, processing external text carries an inherent risk of embedded instructions affecting agent behavior.
[COMMAND_EXECUTION]: The workflow involves extensive file system interactions, including reading and writing to the .memory/activeContext.md file and creating structured records in a references/ directory. These operations are used to maintain project state and organize research outputs.

literature-review